Autonomous Identification And Grasping Of Robotic Flexible Cables Based On Point Cloud

With the rapid development of robot technology,it has become an inevitable trend to replace manual operation with robots in the field of high-intensity and high-risk work.This article takes the research and development of a distribution line maintenance robot system as the background of the project,and focuses on the difficulty of picking and placing weakly rigid and irregularly shaped operating targets by the robot system.The system designed in this paper can realize the automatic recognition and picking and placing of flexible cables based on point cloud.The main work of this thesis is as follows:Aiming at the working environment of the robot and the physical characteristics of the grabbed object,the overall structure of the maintenance robot system is given.Based on the ROS platform,this article proposes a general scheme for autonomously identifying and grabbing lead wires of robot in power distribution lines using point cloud.Based on the analysis of the manipulator's working space and the point cloud of the grasped object,a point cloud preprocessing method combining point cloud filtering and outlier removal is proposed.The point cloud semantic segmentation algorithm based on deep learning is researched.This article intersperses Point SIFT modules into Point Net networks and obtains the trained model through pre-training and fine-tuning.The model is used to achieve semantic segmentation of the target based on point cloud.Experiments show that the algorithm proposed have great advantages in real-time and accuracy.Aiming at the problem that it is difficult to determine the grasping posture of flexible cables with indeterminate shapes,a 3D reconstruction method of leading-wire models based on point cloud is proposed.This thesis proposes an improved K-Means clustering algorithm to divide the point cloud of leading-wire into multiple point cloud clusters and performs the cylindrical fitting on each point cloud cluster.The center of each fitted cylinder is also the discrete points distributed on the center line of the leading-wire.Since the order of discrete points is unknown,a sorting algorithm based on PCA and octree orientation constraints is proposed.After that,the B-spline interpolation algorithm is used to fit the centerline equation of the leading-wire to achieve reconstruction.The validity of the proposed method is verified by experiments,and the effect of 3D reconstruction of the leading-wire can meet the accuracy requirements in subsequent grasping of robots.The autonomous motion planning of the robotic arm to pick and place leading-wire is studied.First considering that the leading-wire is weak rigid,the principle of selecting grabbing points of the leading-wire is proposed.Then through establishing the hand-eye calibration model of the depth camera and the kinematics of the robot,the gripping point is transformed to configuration space.Finally this paper proposes an improved RRT algorithm,including improving the collision detection algorithm involved and changing the sampling strategy,which improved the efficiency of the path planning algorithm and the quality of the planned path.The results of the actual robot gripping experiments show that the real-time performance and path quality of the improved path planning algorithm are greatly improved.It also verifies the feasibility of the method proposed in this paper to automatically grip leading-wire with robot based on point cloud.This method lays a good foundation for outdoor power distribution line maintenance using robot.